ORIGINAL RESEARCH ARTICLE

Prediction of Fetal Darunavir Exposure by Integrating HumanEx-Vivo Placental Transfer and Physiologically BasedPharmacokinetic Modeling

Stein Schalkwijk1,2 • Aaron O. Buaben1 • Jolien J. M. Freriksen2 •

Angela P. Colbers1 • David M. Burger1 • Rick Greupink2 • Frans G. M. Russel2

Published online: 25 July 2017

� The Author(s) 2017. This article is an open access publication

Abstract

Background Fetal antiretroviral exposure is usually

derived from the cord-to-maternal concentration ratio. This

static parameter does not provide information on the

pharmacokinetics in utero, limiting the assessment of a

fetal exposure–effect relationship.

Objective The aim of this study was to incorporate pla-

cental transfer into a pregnancy physiologically based

pharmacokinetic model to simulate and evaluate fetal

darunavir exposure at term.

Methods An existing and validated pregnancy physiolog-

ically based pharmacokinetic model of maternal darunavir/

ritonavir exposure was extended with a feto-placental unit.

To parameterize the model, we determined maternal-to-

fetal and fetal-to-maternal darunavir/ritonavir placental

clearance with an ex-vivo human cotyledon perfusion

model. Simulated maternal and fetal pharmacokinetic

profiles were compared with observed clinical data to

qualify the model for simulation. Next, population fetal

pharmacokinetic profiles were simulated for different

maternal darunavir/ritonavir dosing regimens.

Results An average (±standard deviation) maternal-to-fe-

tal cotyledon clearance of 0.91 ± 0.11 mL/min and fetal-

to-maternal clearance of 1.6 ± 0.3 mL/min was deter-

mined (n = 6 perfusions). Scaled placental transfer was

integrated into the pregnancy physiologically based phar-

macokinetic model. For darunavir 600/100 mg twice a day,

the predicted fetal maximum plasma concentration, trough

concentration, time to maximum plasma concentration, and

half-life were 1.1, 0.57 mg/L, 3, and 21 h, respectively.

This indicates that the fetal population trough concentra-

tion is higher or around the half-maximal effective dar-

unavir concentration for a resistant virus (0.55 mg/L).

Conclusions The results indicate that the population fetal

exposure after oral maternal darunavir dosing is therapeutic

and this may provide benefits to the prevention of mother-

to-child transmission of human immunodeficiency virus.

Moreover, this integrated approach provides a tool to

prevent fetal toxicity or enhance the development of more

selectively targeted fetal drug treatments.Electronic supplementary material The online version of thisarticle (doi:10.1007/s40262-017-0583-8) contains supplementarymaterial, which is available to authorized users.

Stein Schalkwijk and Aaron O. Buaben contributed equally to the

article.

& Frans G. M. Russel

Frans.Russel@radboudumc.nl

1 Department of Pharmacy, Radboud Institute for Health

Sciences, Radboud University Medical Center, Nijmegen,

The Netherlands

2 Department of Pharmacology and Toxicology, Radboud

Institute for Molecular Life Sciences, Radboud University

Medical Center, PO Box 9101, 6500 HB Nijmegen, The

Netherlands

Clin Pharmacokinet (2018) 57:705–716

https://doi.org/10.1007/s40262-017-0583-8

Key Points

Fetal exposure to maternally administered

medication is an important determinant of fetal drug

toxicity or efficacy. Darunavir crosses the placenta

and is frequently used in pregnancy.

We can predict fetal exposure to darunavir, co-

administered with ritonavir, by integrating human

ex-vivo placental transfer and physiologically based

pharmacokinetic modeling.

The placental perfusion setup is a valuable

experimental tool that can be integrated with

physiologically based pharmacokinetic modeling to

simulate fetal drug exposure to darunavir during

pregnancy. Fetal darunavir trough concentration is

just above the half-maximal effective concentration

for wild-type human immunodeficiency virus and

may benefit the prevention of mother-to-child

transmission.

The approach described in this study can be used to

optimize maternal pharmacotherapy to prevent fetal

toxicity or enhance the development of more

selectively targeted fetal drug treatments.

1 Introduction

Many women use medications during the course of their

pregnancy for different reasons [1, 2]. For ethical and

practical considerations, however, pregnant women are

generally excluded from clinical studies [3–5]. Conse-

quently, for the large majority of medications used during

pregnancy, there is inadequate or no information on fetal

exposure and safety. This imposes severe limitations on

drug therapy during pregnancy as placental transfer

potentially poses a threat to fetal well-being. However, it

may allow drugs to be administered to the mother for

therapeutic benefit of the fetus in utero.

Experimentally, fetal drug exposure remains difficult to

quantify, as the fetus and placenta are not readily acces-

sible for sampling until delivery. Sampling of umbilical

cord blood and maternal blood at the time of delivery is an

ethically acceptable and easily accessible method, allowing

the calculation of a cord blood-to-maternal blood concen-

tration ratio. Though a cord blood-to-maternal blood con-

centration ratio is a clinically useful index of relative fetal

drug exposure, there is often large variability between

mother-infant pairs owing to variability in the timing of the

sample collection relative to the last maternal dose and it

does not provide any information on the fetal concentra-

tion–time profile [6, 7]. Ex-vivo dual-side perfusion of a

single human placental cotyledon has proven to be a clin-

ically relevant model for studying placental transport of

various endogenous compounds and xenobiotics [6].

Although this model gives a good estimate of placental

transfer, it does not provide information on in-vivo fetal

pharmacokinetics and exposure.

Physiologically based pharmacokinetic (PBPK) models

may provide a solution. Such models incorporate preg-

nancy-induced changes in various physiological and

anatomical parameters and represent a feasible approach

for appropriate dose optimization in pregnant women [8].

A number of pregnancy PBPK (p-PBPK) models have been

developed to predict the disposition of various drugs in

pregnant women; however, these models are mainly related

to maternal pharmacokinetics [9–12].

It remains a challenge to predict fetal drug exposure

using p-PBPK models that include human placental

transfer parameters. Various animal models have been

used to study placental drug transfer, but data are of poor

translational value because of interspecies variability in

placental structure and hemodynamics [6, 13]. Very few

p-PBPK models include actual human placental transfer

parameters from ex-vivo placenta perfusion experiments

[14, 15]. Such parameters are obtained by using intact

placental tissue with corresponding hemodynamics and

should include active as well as passive drug transfer

processes.

Darunavir administered once daily (QD) or twice daily

(BID), in combination with low-dose ritonavir (darunavir/

ritonavir) is a preferred agent to be used in human

immunodeficiency virus (HIV)-positive pregnant women

[16]. In plasma, darunavir is approximately 94% protein

bound, mainly to a-1-acid glycoprotein (AAG). Bio-

transformation is almost exclusively mediated by cyto-

chrome P450 (CYP) 3A4. Clinically, darunavir is co-

administered with the potent CYP3A4 inhibitor ritonavir,

to reduce darunavir clearance and maintain higher plasma

concentrations throughout the dosing interval [12].

Ritonavir is also known to have potent inhibitory effects

on the efflux drug transporter, P-glycoprotein (P-gp),

which in the intestine, may contribute to the increased

darunavir bioavailability [17]. Recently, we reported on a

whole-body p-PBPK model to describe the maternal

pharmacokinetics of ritonavir-boosted darunavir in preg-

nancy [12]. Darunavir was considered a good candidate

for further development of the p-PBPK model by

including placental transfer and an in-utero compartmen-

tal structure because clinical data on its pharmacokinetics

706 S. Schalkwijk et al.

in pregnancy and cord blood concentrations are available

[18].

In this study, the main focus was to develop a p-PBPK

model to quantitatively predict fetal exposure to the

antiretroviral compound, darunavir, co-administered with

ritonavir. To achieve this, a previously developed p-PBPK

model of darunavir [12] was extended by incorporating a

feto-placental unit into the maternal model. Bi-directional

placental transfer parameters for darunavir were deter-

mined using the ex-vivo, dual-sided, perfused isolated

cotyledon model.

2 Methods

Maternal-to-fetal (MTF) transfer of darunavir has been

previously studied with the ex-vivo cotyledon perfusion

model [19]. In contrast to this study, we included ritonavir

and evaluated MTF and fetal-to-maternal (FTM) placental

darunavir transfer to account for the potential interaction

with placental drug transporters [20]. In brief, placental

transfer parameters for darunavir were determined using

the ex-vivo, dually perfused, isolated cotyledon model.

Next, the p-PBPK model of darunavir developed previ-

ously [12] was extended by a feto-placental unit including

placental transfer based on ex-vivo data using Berkeley

Madonna software (Berkeley Madonna Inc, California).

After model parameterization, simulated maternal and fetal

pharmacokinetic profiles were compared with observed

clinical data.

2.1 Placenta Perfusion

The study was approved by the local Ethics Committee of

Radboud University Medical Center, Nijmegen, the

Netherlands (file number 2014-1397). The experimental

setup and methodology were detailed previously [21]. The

Electronic Supplementary Material provides more details

and specifications on the ex-vivo placenta perfusions and

bioanalyses.

2.2 Modeling

A full-body PBPK model comprising 13 tissue/organ

compartments was built and coded in Berkeley Madonna

syntax, Version 8.3.23.0 (http://www.berkeleymadonna.

com/). The model was largely based on the darunavir/ri-

tonavir p-PBPK model developed in Simcyp� Version 13

release 2 (Simcyp Limited, a Certara company, Sheffield,

UK), and described previously for maternal darunavir

pharmacokinetics [12]. All Berkeley Madonna codes are

available upon request. No model fitting and/or parameter

estimation was performed in the current study.

2.3 Physiologically Based Pharmacokinetic Model

Development

Human physiological parameters were obtained from liter-

ature as well as from virtual populations of healthy volun-

teers and pregnant women implemented in Simcyp�.

Physicochemical and in-vitro pharmacokinetic parameters

of darunavir and ritonavir were obtained from literature [12].

First, a PBPK model was developed for darunavir in non-

pregnant women for oral administration. The model was

validated by comparing simulations with observed in-vivo PK

profiles obtained after oral administration of standard dosages.

Once the PBPK model was evaluated in the non-pregnant

population, all drug-specific parameters were fixed and

pregnancy-specific changes in physiological parameters were

incorporated. Model performance was verified by comparing

simulations with observed concentrations in pregnant women

[22] as well as with the previously developed p-PBPK model

of Colbers et al. [12]. The p-PBPK model was extended by

incorporation of a feto-placental unit (Fig. 1). Simulated fetal

plasma darunavir concentrations were compared with repor-

ted cord blood concentrations. Only population parameters

Fig. 1 Final pregnancy physiologically based pharmacokinetic

model including the feto-placental unit. EHR enterohepatic recircu-

lation, fa fraction absorbed, CLaf estimated clearance from amniotic

fluid to fetal blood, CLfa estimated clearance from fetal blood to

amniotic fluid, CLhepatic hepatic clearance, CLfm fetal-to-maternal

placental clearance, CLmf maternal-to-fetal placental clearance, Q re-

spective tissue/organ blood flow. The subscripts are denoted as

follows: ad adipose, bo bone, br brain, he heart, ki kidney, lu lung, mu

muscle, sk skin, sp spleen, gu gut, li liver, RoB rest of fetal body

Prediction of Fetal Darunavir Exposure 707

were considered and individual variability was not included.

We conducted sensitivity analyses to assess the impact of

changes in feto-placental-related parameters on fetal plasma

concentration–time profiles.

Additionally, we explored the influence of different

darunavir/ritonavir dosing regimens on fetal exposure. This

was mainly done to illustrate its usefulness for assessing

exposure–response relations. For this purpose, we assumed

a minimum effective fetal plasma Ctrough of 0.55 mg/L.

This is based on the half-maximal effective darunavir

concentration for resistant HIV, a target frequently used in

therapeutic drug monitoring [23].

3 Results

3.1 Ex-vivo Perfusion of Human Placental

Cotyledon

A total of 15 placentas were collected from elective Cae-

sarean sections (n = 8) and uncomplicated vaginal

deliveries (n = 7). Six perfusions fulfilled the quality cri-

teria for successful perfusion. On average, perfusion was

started within 45 min of delivery; which is consistent with

previous studies [24]. We conducted a perfusion experi-

ment without placental tissue and found no indications for

system adherence of darunavir and ritonavir (data not

shown).

The concentration–time profiles of the FTM and MTF

perfusion experiments are shown in Fig. 2a, c. Assuming

that the amounts in the maternal, fetal, and placental

compartments at each time point sum up to the total

amount (100%) added at t = 0, the placental content was

calculated (Fig. 2b, d). This was based on the absence of

system adherence and the presumed absence of darunavir

biotransformation in the perfused cotyledon. Sampling

from the cotyledon over the course of the experiment was

not feasible as it would have caused leakage. After each

experiment, a cotyledon tissue sample was taken for bio-

analyses (Fig. 2a, c). In the mass balance calculation, we

accounted for volume and darunavir loss because of sam-

pling. Darunavir did not accumulate in placental tissue. Bi-

Fig. 2 Darunavir concentration–time and mass-balance profiles in

ex-vivo placenta perfusion. Placental transfer of darunavir in the

presence of ritonavir over 150 or 180 min of perfusion. a Maternal-to-

fetal transfer; data represent darunavir concentrations in the maternal

reservoir (filled circles), maternal outflow (open circles), and fetal

outflow (open triangles) after administration of darunavir/ritonavir (6/

0.5 mg/L) to the maternal circulation at t = 0 min. b Maternal-to-

fetal transfer; accumulation, and mass balance calculation (expressed

as percentage of the initial maternal reservoir concentration) showing

that tissue concentrations reach steady state at t = 60 min. c Fetal-to-

maternal transfer; data represent darunavir concentrations in the fetal

reservoir (filled triangles), fetal outflow (open triangles), and

maternal outflow (open circles) after administration of darunavir/

ritonavir (6/0.5 mg/L) to the fetal circulation at t = 0 min. Placental

tissue concentrations at the end of the perfusion are shown as stars.

d Fetal-to-maternal transfer; accumulation, and mass balance calcu-

lation (expressed as percentage of the initial fetal reservoir concen-

tration) showing that tissue concentrations reach steady state at

t = 60 min

708 S. Schalkwijk et al.

directional clearances were determined based on linear

regression analysis of natural log-transformed darunavir

concentrations in the closed reservoir after 60 min.

The MTF and FTM cotyledon clearances were calcu-

lated as follows:

Ex-vivo cotyledon clearance

CLcot ¼ ke � V; ð1Þ

where CLcot represents the mean cotyledon clearance val-

ues, V is the volume of the perfusion reservoir, and ke is a

placental elimination constant derived from the slope of the

natural log concentration–time profile (MTF, R2 = 0.994;

FTM, R2 = 0.976).

3.2 Development and Verification of a Darunavir/

Ritonavir Physiologically Based

Pharmacokinetic Model Model in Non-Pregnant

Subjects

Time-based differential equations used to describe changes

in drug concentrations in various organ compartments were

as follows:

Maternal non-eliminating tissue compartments (perfu-

sion limited)

Vtissue

dCtissue

dt¼ Qt � Carterial �

Ctissue

Kpt

� BP

� �; ð2Þ

Well-stirred liver model

Vliver

dCliver

dt¼ Qha � Carterial þ Qsp �

Csp

Kpsp

� BP

!

þ Qgu �Cgu

Kpgu

� BP

!

� Qli �Cli

Kpli

� BP

� �� CLintmet � fuli � Cli;

ð3Þ

where V, Cli, Carterial, Csp, and Cgu are volumes, drug

concentrations in liver, arterial blood, spleen, and gut,

respectively. The parameter Q denotes the respective blood

flow in or out of the organ/tissue. CLhepatic, BP, and Kp are

total hepatic clearance, blood-to-plasma ratio, and tissue-

to-plasma partition coefficient, respectively. The subscripts

ha, sp, gu, li, and t denote the hepatic artery, spleen, gut,

liver, and tissue, respectively.

Absorption was defined by a one-compartment model

with first-order absorption rate, ka. Drug administered in

the dosing compartment (gut lumen) is absorbed into the

gut wall, from which the drug is released into the portal

circulation.

Perfusion-limited distribution kinetics was used for all

non-eliminating tissues; hence, passive diffusion from

plasma into tissue and homogenous distribution across the

tissue mass was assumed. Tissue-to-plasma partition

coefficients (Kp) used in this model were previously pre-

dicted by us in Simcyp�, using the Rodgers and Rowland

method [12, 25]. Because a well-stirred liver model (per-

fusion-limited) resulted in substantial overestimation of

darunavir exposure (in the absence of ritonavir) [12], a

permeability-limited liver model including active uptake of

the drug into the liver was used for the previous p-PBPK

model of darunavir in pregnancy. In addition, empirical

clinical data exist suggesting permeability-limited distri-

bution into the eliminating tissues [26]. However, full

parameterization of such a model requires data on activity

and abundance of the drug transporters involved, which are

currently not available. Because the scope of this work was

mainly related to fetal pharmacokinetics, we used a sim-

plified and refined well-stirred liver model against the

background of their permeability-limited liver model,

which includes active uptake of the drug into the liver. In

refining the well-stirred liver model, we assumed that drug

exchange in the liver is primarily driven by passive diffu-

sion between plasma and the interstitial space; and that

there is active transporter-mediated uptake of drug into the

cell, in addition to passive diffusion. This leads to

increased intracellular darunavir availability and thus,

biotransformation.

Based on steady-state conditions, we derived the fol-

lowing equation:

CuLi ¼ Cup � UF; ð4Þ

where Cup is the unbound plasma concentration of drug,

CuLi is the unbound drug concentration in the liver, and UF

is an uptake ratio. The uptake parameter, UF, was esti-

mated by fitting model simulations against the observed

clinical data (darunavir/ritonavir 600/100 mg in non-preg-

nant adults) presented by Rittweger et al. (27). An optimal

UF of 1.2 was estimated against the background of free

plasma-to-tissue concentration ratio predicted in Simcyp�

[12] and observed in vivo [28].

Systemic clearance of darunavir was considered to

mainly occur via the liver; therefore, renal clearance was

not included in the model [27]. Because enterohepatic

recirculation of darunavir was a major determinant in

modeling the interaction of darunavir with ritonavir in the

existing p-PBPK model, and was suggested in a previous

clinical study [29], an enterohepatic recirculation parame-

ter was added to the current model as an empirical solution

by visually fitting the simulated PK profile to the observed

clinical data (non-pregnant). The enterohepatic recircula-

tion was defined as a constant fraction of the modeled

darunavir dose available for re-absorption from the gut

lumen. This is based on the assumption that a small

Prediction of Fetal Darunavir Exposure 709

fraction of darunavir amount is excreted through biliary

clearance to the gut lumen.

The darunavir-ritonavir drug–drug interaction model

used includes both competitive and mechanism-based

inhibition of CYP3A4 activity. A static ritonavir plasma

concentration was used for inhibition. Different population

ritonavir unbound steady-state Ctrough were used for BID

and QD dosing, and for the pregnant and non-pregnant

simulations [22]. Predicted PK profiles were in line with

observed clinical data as well as with the simulated PK

profiles by Colbers et al. (Fig. 3).

3.3 Development and Verification of a Darunavir/

Ritonavir Physiologically Based

Pharmacokinetic Model Model in Pregnant

Subjects

After successful simulation of PK profiles in non-pregnant

subjects, physiological parameters were modified to reflect

changes in pregnancy, while keeping all drug-specific

parameters constant. Physiological and metabolic changes

(e.g., body weight and CYP enzyme activity) with gesta-

tional age were implemented using data and algorithms

described previously [30]. Variations in protein binding

during pregnancy as well as in the fetus at term were

implemented in the model based on algorithms from

Simcyp�, assuming the absence of changes in protein

binding kinetics. Further details on model parameterization

were described previously [12].

The PK profile of darunavir was simulated in pregnant

women (gestational age = 38 weeks) for a steady-state

situation of the clinical dosage regimens 600/100 mg BID

as well as 800/100 mg QD. Visual inspection of the curves

showed that the simulated PK profile was in line with

observed clinical data as well as simulated PK profiles of

Colbers et al. [12] (Fig. 4a, b). Based on this verification,

we considered the model suitable for extension with a feto-

placental unit to simulate fetal PK profiles.

3.4 Incorporation of a Feto-Placental Unit

3.4.1 Scaling Ex-Vivo Cotyledon Darunavir Clearance

to In-Vivo Placental Clearance

The ex-vivo cotyledon clearance of darunavir determined

from the perfusion experiments was scaled to in-vivo pla-

cental clearance. Because darunavir displays a substantial

degree of protein binding (94%) in plasma, it was neces-

sary to assess the free darunavir fraction in the perfusion

samples (containing 30 g/L albumin) to calculate an

unbound cotyledon clearance. The mean, concentration-

independent, protein-free fraction of darunavir in the per-

fusion medium was 16% ± 0.5% (n = 5). Second, the

unbound cotyledon clearance was extrapolated to in-vivo

cotyledon clearance, taking into account the respective

protein-free fractions in maternal and fetal blood and the

average number of cotyledons in a placenta (Table 1).

The equations used are as follows:Scaled placental

clearance

CLmf ¼ fup � CLcotmf= fuperf

� �� Ncot;

CLfm ¼ fupF � CLcotfm= fuperf

� �� Ncot;

ð5Þ

where Ncot denotes the average number of cotyledons per

placenta, CLcot, mf is the MTF cotyledon clearance,

CLcot, fm is the FTM cotyledon clearance, fuperf is the

protein-free fraction in the perfusion medium, CLmf is in-

vivo MTF placental clearance, CLfm is in-vivo FTM pla-

cental clearance, fup is the fraction unbound in maternal

plasma, and fupF is the fraction unbound in fetal plasma.

Fig. 3 Simulation of maternal (non-pregnant) darunavir concentra-

tion–time profiles. Simulation of maternal (non-pregnant) darunavir

pharmacokinetic profiles at steady state for darunavir/ritonavir (DRV/

r) dosing regimens of a 600/100 mg twice daily (BID) and b 800/

100 mg once daily (QD). These simulations are compared with

previous physiologically based pharmacokinetic simulations [12], as

well as observed clinical data [22]. conc. concentration

710 S. Schalkwijk et al.

Variations in protein binding in maternal and fetal blood

were taken into account in this model. Fraction of unbound

drug in maternal and fetal blood was predicted from

algorithms available in Simcyp�, assuming the absence of

changes in binding kinetics for AAG, the predominant

binding protein for darunavir [12, 27]. However, fetal

plasma AAG levels are appreciably lower than in the

maternal circulation (fetal:maternal AAG ratio = 0.37)

[6]. Predicting the darunavir fraction unbound in fetal

plasma based on AAG as the primary binding protein and

assuming an average fetal AAG level at term of 0.17 g/L

[6], resulted in a predicted fupF of 0.22. This value is higher

than the unbound fraction (fuperf) of 0.16 determined in the

experimental placenta perfusion medium in the absence of

AAG, but in the presence of albumin (30 g/L). This indi-

cates that in vivo, albumin (33.5 g/L) may be more

important in binding darunavir in the fetal circulation [31]

relative to the low AAG levels. Therefore, predictions of

the darunavir fraction unbound in the fetal circulation were

based on albumin levels. The predicted fraction unbound in

maternal and fetal blood were 0.12 and 0.27, respectively.

Fig. 4 Simulation of maternal and fetal darunavir concentration–time

profiles. Simulation of maternal and fetal plasma darunavir pharma-

cokinetic profiles at steady state for darunavir/ritonavir (DRV/r)

dosing regimens of a 600/100 mg twice daily (BID) and b 800/

100 mg once daily (QD). Simulated fetal plasma darunavir

concentrations compared with observed cord blood concentrations

for the maternal doses of c 600/100 mg BID and d 800/100 mg OD.

These simulations are compared with previous pregnancy physiolog-

ically based pharmacokinetic simulations [12], as well as observed

clinical data [18, 22]. conc. concentration, GA gestational age

Table 1 In-vitro and physiological parameters for the feto-placental

unit

Parameter; gestational age: 38 weeks Value References

Fetal cardiac output, Qfco (L/h) 85.5 [32]

Fetal weight (kg) 3.56 [30]

Amniotic fluid volume (L) 0.86 [14, 30]

Fetal blood volume, VbloodFetal (L) 0.24 [44]

DRV CLaf (L/h) 0.04 Estimated

DRV CLfa (L/h) 0.08 Estimated

DRV CLcot, mf (mL/min) 0.91 ± 0.11 Determined

DRV CLcot, mf (mL/min) 1.6 ± 0.3 Determined

Ncot 30 [45]

DRV fuperf 0.16 Determined

Hct 50% [31]

CLaf amniotic fluid-to-fetal blood clearance, CLfa fetal blood-to-am-

niotic fluid clearance, CLcot, fm fetal-to-maternal cotyledon clearance,

CLcot, mf maternal-to-fetal cotyledon clearance, fuperf free fraction of

darunavir in perfusion medium, Hct hematocrit, Ncot average number

of cotyledons per placenta

Prediction of Fetal Darunavir Exposure 711

3.4.2 Fetal Physiologically Based Pharmacokinetics

of Darunavir/Ritonavir

The placenta was considered as a barrier between maternal

and fetal blood. The fetal compartment was split into fetal

blood and rest of fetal body (Fig. 1). During late pregnancy,

20% of fetal cardiac output is distributed to the placenta, with

the remaining 80% distributed to the rest of the fetal body

(QRoB; Fig. 1) [32]. Fetal tissues were lumped into one

compartment. Tissue partitioning was based on reported

adult Vss, assuming linearity between volume of distribution

and body volume [12]. Although there may be differences in

fetal Vss and adult Vss, no data are currently available to

address such differences. Moreover, within-species Vss is

usually well predicted by allometry [33].

Darunavir is also reported to distribute into amniotic

fluid up to *25% of maternal concentrations [34].

Although these data are limited, we assumed slow mass

transfer (Table 1) from and into amniotic fluid to reach a

steady-state amniotic concentration of about 25%. Mass

transfer was assumed to be relatively slow compared with

placental clearance because the processes of excretion and

absorption into and from amniotic fluid are likely to be

much slower than placental clearance.

Because of low abundance and activity of CYP3A enzymes

in the placenta [35], it was assumed in this study that there was

negligible placental metabolism of darunavir. Furthermore, in

the presence of ritonavir, we would expect any remaining

placental CYP metabolism to be completely inhibited. In

addition, the relevance of fetal hepatic metabolism was

assumed to be negligible and therefore not taken into account.

Although large variability is reported, relative CYP3A4 gene

expression was found to be 40,000 times higher in adults than

in fetuses, whereas CYP3A7 expression was 5500 higher in

fetuses than in adults [36]. It cannot be excluded that in utero

other CYP450 enzymes (e.g., CYP3A7) may be involved in

the biotransformation of darunavir. More research is needed to

generate adequate data to be able to incorporate fetal hepatic

metabolism into p-PBPK models in a mechanistic manner.

Darunavir disposition in the feto-placental unit was

described as follows:

Fetal blood

VbloodF

dCbloodF

dt¼ CLmf � Cart � CLfm � CbloodF

þ QrobF �CrobF

KprobF

� BPF � CbloodF

� �

� CbloodF � CLfa þ CLaf � Camf ; ð6Þ

Fetal body

VrobF

dCrobF

dt¼ QrobF � CbloodF �

CrobF

KprobF

� BPF

� �; ð7Þ

Amniotic fluid

Vamf

dCamf

dt¼ CbloodF � CLfa � CLaf � Camf ; ð8Þ

where CLmf denotes MTF placental clearance, CLfm is the

FTM placental clearance, BPF is the fetal blood-to-plasma

ratio, QrobF is the blood flow to the fetal body, V is the

tissue volumes, C is the concentration, Q is the blood flow,

KprobF is the fetal tissue partition coefficient, CLfa is the

estimated clearance from fetal blood to amniotic fluid, CLaf

is the estimated clearance from amniotic fluid to fetal

blood, and robF is the rest of the fetal body. The subscripts

amf, bloodF, and robF denote amniotic fluid, fetal blood,

and rest of fetal body, respectively.

The parameters used to model the feto-placental unit are

listed in Table 1. Simulated fetal plasma darunavir con-

centrations were comparable to observed cord blood con-

centrations (Fig. 4). From the observed clinical data, the

median (range) ratio for darunavir cord plasma/maternal

plasma was 0.2 (0.0–0.8) [22]. The simulated population

fetal plasma-to-maternal plasma concentration ratio over a

dosing interval is 0.30 (0.16–0.37).

3.5 Sensitivity Analysis

The results of the sensitivity analysis are shown in Fig. 5.

Changes in fetal darunavir tissue partitioning affected the

amplitude of the steady-state concentration-time profiles,

which corresponds to changes in Vss. The steady-state fetal

plasma concentration–time profiles were not affected by

the extent of darunavir clearance from the amniotic fluid to

fetal blood, mainly because clearance into and from

amniotic fluid was substantially lower than placental

clearance and partitioning into fetal tissue. This indicates

that, as previously observed [14], within physiologically

plausible ranges poorly informed assumptions made on

partitioning into amniotic fluid do not influence the out-

come of interest, i.e., fetal plasma concentrations. With

regard to changes in fraction unbound in fetal plasma and

MTF cotyledon clearance (parameterized proportionally),

both substantially impact the fetal plasma concentrations.

Therefore, precise (experimental) estimation and external

validity of these parameters is essential.

3.6 Exploratory Simulations of Fetal Exposure

to Darunavir at Different Doses

Based on the assumed exposure–effect relationship for

fetal antiviral activity, we conducted simulations to deter-

mine the optimal maternal dosing regimen with respect to

fetal treatment outcomes. We simulated a dose range for

both QD and BID maternal dosing (Fig. 6). These simu-

lations indicate that the standard dosing regimens result in

712 S. Schalkwijk et al.

therapeutic exposure to darunavir. However, if HIV-posi-

tive women have developed (multiple) resistance to pro-

tease inhibitors, possibly even higher dosing regimens are

optimal with regard to fetal benefit, especially in the case

of high-risk situations, such as maternal HIV breakthrough

at term.

4 Discussion

We developed a p-PBPK model for fetal drug exposure

during pregnancy that allowed us to predict the fetal

pharmacokinetics of ritonavir-boosted darunavir, at term.

In this study we introduced a feto-placental unit in the

model using bidirectional placental clearance parameters

determined separately from an ex-vivo human cotyledon

perfusion set-up.

To date, few p-PBPK models have simulated human

fetal PK profiles by integrating ex-vivo human placental

transfer parameters within a p-PBPK model. Two studies

report placental transfer of two other antiretroviral treat-

ments based on placental diffusion and placental elimina-

tion, using closed system, ex-vivo placental perfusion

experiments and non-linear compartmental modeling

[14, 15]. Although the derived physiological parameters

can be used to simulate fetal pharmacokinetics, it is chal-

lenging to precisely identify all the required parameters

using a one-way, closed-system perfusion setup [37].

The approach used in this study provides an alternative

to integrating placental transfer into p-PBPK models.

Bidirectional transfer parameters were determined sepa-

rately, by performing MTF and FTM perfusion experi-

ments. The placental clearances estimated from the current

ex-vivo experiments can be considered as whole organ

clearances and allowed inclusion of placental transfer as a

single barrier, rather than a compartmental structure with

corresponding flows, partitioning coefficients, and tissue

volumes. In terms of future development of more mecha-

nistic compartmental placental distribution models, for

instance for in-vitro-to-in-vivo extrapolation of placental

transfer data from passive permeability and placental active

transport studies, this approach in PBPK modeling may

provide a valuable starting point for verification steps.

With regard to the verification of simulated fetal expo-

sure to drugs in general, hardly any information is avail-

able. Preclinical studies indicate substantial fetal darunavir

exposure in rats, but extrapolation of these data to human

pregnancy is challenging [38]. Umbilical cord-to-plasma

concentration ratios in pregnant women taking darunavir/

ritonavir were reported [22]. Such data allow a rough

Fig. 5 Sensitivity analysis of feto-placental parameters. Sensitivity

analysis of feto-placental parameters with respect to fetal plasma

darunavir concentration–time profiles at steady state for darunavir/

ritonavir dosing regimens of 600/100 mg twice daily. The parameters

of interest include a fetal tissue partition coefficient), b amniotic

fluid-to-fetal blood clearance, c fraction unbound in fetal plasma, and

d maternal-to-fetal cotyledon clearance. conc. concentration

Prediction of Fetal Darunavir Exposure 713

verification of the model by comparing simulated profiles

with actual observed cord blood concentrations, as this is

the closest proxy for fetal exposure clinically and ethically

available [7]. The simulated fetal concentration–time pro-

file corresponds with the range of observed cord blood

concentrations, indicating that the developed darunavir

PBPK model provides a good approximation of the fetal

PK profile. The observed fetal plasma concentrations fol-

lowing darunavir 600 mg BID (Fig. 4c) show high vari-

ability. This can be misinterpreted as accumulation of the

drug, which is unlikely under steady-state conditions.

Caution is needed when interpreting concentration–time

profiles based on one observation per subject, with a lim-

ited amount of subjects, which is also the reason why data-

driven population pharmacokinetics is not an option in this

case. Consequently, we looked at the range rather than the

shape of these observed concentrations.

The observed FTM placental clearance in this study was

higher than MTF placental clearance. One explanation

could be that darunavir is a substrate for P-gp/ABCB1, an

efflux transporter located at the apical surface of placental

syncytiotrophoblast membrane. Darunavir efflux from tro-

phoblasts into the maternal plasma could therefore result in

relatively low MTF clearance. However, the experiment

was conducted in the presence of ritonavir, which is known

to inhibit P-gp, while also, P-gp functional expression is

expected to be low in the term placentas used in this study

because of reported gestational changes [39] Overall, the

role of transporters in regulating passage across the pla-

centa remains unclear [40, 41]. Because the system was

operated under sink conditions, the higher FTM clearance

can also be related by the higher flow rate in the maternal

compartment, which maintains a steeper concentration

gradient. Nevertheless, it is of note that the placental

clearance parameters determined from this experimental

set-up are based on physiological flows.

Because term placentas were used, the findings of this

study are limited mainly to drug administration at term.

Additionally, inferences on the impact of ritonavir con-

centrations on the darunavir placental transfer cannot be

made. In general, human ex-vivo cotyledon perfusion

experiments are limited to small numbers of term placen-

tas. Studies over a large concentration range or with pla-

centas from earlier stages of pregnancy are time consuming

and generally less feasible [6]. More data on placental

CYP450 expression, changes in hemodynamics, tissue

composition, and active uptake and efflux transport may

provide a better mechanistic basis for the description and

prediction of the placental transfer processes. Placental

transfer models could be developed that allow for the

simulation of fetal PK profiles in early pregnancy and fetal

exposure-response relationships, including drug–drug

interactions.

Another interesting finding in this study was the obser-

vation that cotyledon clearance did not correlate in a linear

manner with cotyledon weight. This is consistent with data

from a previous study, possibly indicating that tissue vol-

ume is a poor indicator of membrane surface area [42].

Based on this observation, the cotyledon clearance was

scaled per cotyledon assuming that with multiple perfusion

experiments, the population clearance per cotyledon is

approximated. Moreover, cutting the perfused part from the

remainder of the placenta is not very precise; therefore,

cotyledon weight was inappropriate to use for further

calculations.

With regard to the prevention of mother-to-child trans-

mission, vertical transmission of HIV from mothers with a

low or undetectable viral load can occur [43]. Successful

quantitative simulations of fetal exposure to antiretroviral

treatments provide a good basis for administering

antiretroviral agents for pre-exposure prophylaxis of the

fetus, but the actual clinical relevance is still under debate

[43]. Nevertheless, as shown by this study, p-PBPK models

can be employed as tools to optimize maternal pharma-

cotherapy to prevent fetal toxicity or enhance the devel-

opment of more selectively targeted, fetal drug treatments.

Fig. 6 Simulated fetal darunavir concentration-time profiles for

several darunavir dosing regimens. Simulated fetal darunavir con-

centration–time profiles for several maternal darunavir doses for

a once-daily (QD) dosing and b twice-daily (BID) dosing. The gray

dashed lines represent the therapeutic target trough concentration.

conc. concentration, DRV/r darunavir/ritonavir

714 S. Schalkwijk et al.

5 Conclusion

A p-PBPK model was developed to quantitatively predict

fetal exposure at term to the protease inhibitor darunavir

co-administered with ritonavir. By extending the existing

maternal PBPK model with a feto-placental unit, in-vitro-

to-in-vivo extrapolation was performed. If applied appro-

priately, the placental perfusion set-up is a valuable

experimental tool that can be integrated with PBPK mod-

eling to simulate fetal drug exposure during pregnancy.

Acknowledgements We thank all the women who donated placentas

and the midwives who collected them. We thank Gerard Zijderveld

for inclusion of participants, Nielka van Erp for supervision of the

bioanalyses, and Noor van Ewijk-Beneken Kolmer for her help with

the bioanalyses. We also are grateful for Khaled Abduljalil’s com-

ments on an earlier version of the manuscript and Certara for the

Simcyp Grant and Partnership Scheme grant allowing further research

on this topic.

Compliance with Ethical Standards

Funding This study was supported by Health * Holland, top sector

Life Sciences & Health.

Conflict of interest Stein Schalkwijk, Aaron O. Buaben, Jolien J.M.

Freriksen, Angela P. Colbers, David M. Burger, Rick Greupink, and

Frans G.M. Russel have no conflicts of interest directly related to this

study.

Open Access This article is distributed under the terms of the

Creative Commons Attribution-NonCommercial 4.0 International

License (http://creativecommons.org/licenses/by-nc/4.0/), which per-

mits any noncommercial use, distribution, and reproduction in any

medium, provided you give appropriate credit to the original

author(s) and the source, provide a link to the Creative Commons

license, and indicate if changes were made.

References

1. Pisa FE, Casetta A, Clagnan E, Michelesio E, Vecchi Brumatti L,

Barbone F. Medication use during pregnancy, gestational age and

date of delivery: agreement between maternal self-reports and

health database information in a cohort. BMC Pregnancy Child-

birth. 2015;15:310.

2. Lupattelli A, Spigset O, Twigg MJ, Zagorodnikova K, Mardby

AC, Moretti ME, et al. Medication use in pregnancy: a cross-

sectional, multinational web-based study. BMJ Open.

2014;4(2):e004365.

3. Blehar MC, Spong C, Grady C, Goldkind SF, Sahin L, Clayton

JA. Enrolling pregnant women: issues in clinical research.

Womens Health Issues. 2013;23(1):e39–45.

4. Shields KE, Lyerly AD. Exclusion of pregnant women from

industry-sponsored clinical trials. Obstet Gynecol.

2013;122(5):1077–81.

5. Mitchell AA, Gilboa SM, Werler MM, Kelley KE, Louik C,

Hernandez-Diaz S, et al. Medication use during pregnancy, with

particular focus on prescription drugs: 1976–2008. Am J Obstet

Gynecol. 2011;205(1):51 e1–8.

6. Hutson JR, Garcia-Bournissen F, Davis A, Koren G. The human

placental perfusion model: a systematic review and development

of a model to predict in vivo transfer of therapeutic drugs. Clin

Pharmacol Ther. 2011;90(1):67–76.

7. Zhang Z, Unadkat JD. Verification of a maternal-fetal physio-

logically based pharmacokinetic model for passive placental

permeability drugs. Drug Metab Dispos. 2017;45(8):939–46.

8. Lu G, Abduljalil K, Jamei M, Johnson TN, Soltani H, Rostami-

Hodjegan A. Physiologically-based pharmacokinetic (PBPK)

models for assessing the kinetics of xenobiotics during preg-

nancy: achievements and shortcomings. Curr Drug Metab.

2012;13(6):695–720.

9. Gaohua L, Abduljalil K, Jamei M, Johnson TN, Rostami-Hod-

jegan A. A pregnancy physiologically based pharmacokinetic (p-

PBPK) model for disposition of drugs metabolized by CYP1A2,

CYP2D6 and CYP3A4. Br J Clin Pharmacol. 2012;74(5):873–85.

10. Ke AB, Nallani SC, Zhao P, Rostami-Hodjegan A, Isoherranen

N, Unadkat JD. A physiologically based pharmacokinetic model

to predict disposition of CYP2D6 and CYP1A2 metabolized

drugs in pregnant women. Drug Metab Dispos.

2013;41(4):801–13.

11. Ke AB, Nallani SC, Zhao P, Rostami-Hodjegan A, Unadkat JD.

Expansion of a PBPK model to predict disposition in pregnant

women of drugs cleared via multiple CYP enzymes, including

CYP2B6, CYP2C9 and CYP2C19. Br J Clin Pharmacol.

2014;77(3):554–70.

12. Colbers A, Greupink R, Litjens C, Burger D, Russel FG. Physi-

ologically based modelling of darunavir/ritonavir pharmacoki-

netics during pregnancy. Clin Pharmacokinet.

2016;55(3):381–96.

13. Carter AM. Animal models of human placentation: a review.

Placenta. 2007;28(Suppl. A):S41–7.

14. De Sousa Mendes M, Hirt D, Vinot C, Valade E, Lui G, Pressiat

C, et al. Prediction of human fetal pharmacokinetics using

ex vivo human placenta perfusion studies and physiologically

based models. Br J Clin Pharmacol. 2016;81(4):646–57.

15. De Sousa Mendes M, Lui G, Zheng Y, Pressiat C, Hirt D, Valade

E, et al. A physiologically-based pharmacokinetic model to pre-

dict human fetal exposure for a drug metabolized by several

CYP450 pathways. Clin Pharmacokinet. 2017;56(5):537–50.

16. DHHS. Recommendations for use of antiretroviral drugs in

pregnant HIV-1-infected women for maternal health and inter-

ventions to reduce perinatal HIV transmission in the United

States [updated August 6, 2015]. Available from: http://aidsinfo.

nih.gov/contentfiles/lvguidelines/perinatalgl.pdf. Accessed 12

July 2017.

17. Holmstock N, Annaert P, Augustijns P. Boosting of HIV protease

inhibitors by ritonavir in the intestine: the relative role of cyto-

chrome P450 and P-glycoprotein inhibition based on Caco-2

monolayers versus in situ intestinal perfusion in mice. Drug

Metab Dispos. 2012;40(8):1473–7.

18. Stek A, Best BM, Wang J, Capparelli EV, Burchett SK, Kre-

itchmann R, et al. Pharmacokinetics of once versus twice daily

darunavir in pregnant HIV-infected women. J Acquir Immune

Defic Syndr. 2015;70(1):33–41.

19. Mandelbrot L, Duro D, Belissa E, Peytavin G. Placental transfer

of darunavir in an ex vivo human cotyledon perfusion model.

Antimicrob Agents Chemother. 2014;58(9):5617–20.

20. US Food and Drug Administration. Center for Drug Evaluation

and Research: pharmacology review. Maryland: Silver spring;

2009.

21. Schalkwijk S, Greupink R, Colbers AP, Wouterse AC, Verweij

VG, van Drongelen J, et al. Placental transfer of the HIV inte-

grase inhibitor dolutegravir in an ex vivo human cotyledon per-

fusion model. J Antimicrob Chemother. 2016;71(2):480–3.

Prediction of Fetal Darunavir Exposure 715

22. Colbers A, Molto J, Ivanovic J, Kabeya K, Hawkins D, Gingel-

maier A, et al. Pharmacokinetics of total and unbound darunavir

in HIV-1-infected pregnant women. J Antimicrob Chemother.

2015;70(2):534–42.

23. US Food and Drug Administration. Prezista label information.

Maryland: Silver spring; 2014.

24. Mathiesen L, Mose T, Morck TJ, Nielsen JK, Nielsen LK,

Maroun LL, et al. Quality assessment of a placental perfusion

protocol. Reprod Toxicol. 2010;30(1):138–46.

25. Toh S, Mitchell AA, Louik C, Werler MM, Chambers CD,

Hernandez-Diaz S. Antidepressant use during pregnancy and the

risk of preterm delivery and fetal growth restriction. J Clin Psy-

chopharmacol. 2009;29(6):555–60.

26. Yoshikado T, Maeda K, Furihata S, Terashima H, Nakayama T,

Ishigame K, et al. A clinical cassette dosing study for evaluating

the contribution of hepatic OATPs and CYP3A to drug-drug

interactions. Pharm Res. 2017;34(8):1570–83.

27. Rittweger M, Arasteh K. Clinical pharmacokinetics of darunavir.

Clin Pharmacokinet. 2007;46(9):739–56.

28. EMA. Prezista; summary of product characteristics 2014.

Available from: http://www.ema.europa.eu/docs/en_GB/

document_library/EPAR_-_Product_Information/human/000707/

WC500041756.pdf. Accessed 12 July 2017.

29. Ter Heine R, Mulder JW, van Gorp EC, Wagenaar JF, Beijnen

JH, Huitema AD. Intracellular and plasma steady-state pharma-

cokinetics of raltegravir, darunavir, etravirine and ritonavir in

heavily pre-treated HIV-infected patients. Br J Clin Pharmacol.

2010;69(5):475–83.

30. Abduljalil K, Furness P, Johnson TN, Rostami-Hodjegan A,

Soltani H. Anatomical, physiological and metabolic changes with

gestational age during normal pregnancy: a database for param-

eters required in physiologically based pharmacokinetic mod-

elling. Clin Pharmacokinet. 2012;51(6):365–96.

31. van den Akker CH, Schierbeek H, Rietveld T, Vermes A,

Duvekot JJ, Steegers EA, et al. Human fetal albumin synthesis

rates during different periods of gestation. Am J Clin Nutr.

2008;88(4):997–1003.

32. Kiserud T, Ebbing C, Kessler J, Rasmussen S. Fetal cardiac

output, distribution to the placenta and impact of placental

compromise. Ultrasound Obstet Gynecol. 2006;28(2):126–36.

33. Anderson BJ, Holford NH. Mechanism-based concepts of size

and maturity in pharmacokinetics. Annu Rev Pharmacol Toxicol.

2008;48:303–32.

34. Else LJ, Taylor S, Back DJ, Khoo SH. Pharmacokinetics of

antiretroviral drugs in anatomical sanctuary sites: the fetal com-

partment (placenta and amniotic fluid). Antivir Ther.

2011;16(8):1139–47.

35. Hakkola J, Pelkonen O, Pasanen M, Raunio H. Xenobiotic-me-

tabolizing cytochrome P450 enzymes in the human feto-placental

unit: role in intrauterine toxicity. Crit Rev Toxicol.

1998;28(1):35–72.

36. Betts S, Bjorkhem-Bergman L, Rane A, Ekstrom L. Expression

of CYP3A4 and CYP3A7 in human foetal tissues and its corre-

lation with nuclear receptors. Basic Clin Pharmacol Toxicol.

2015;117(4):261–6.

37. Mould DR, Upton RN. Basic concepts in population modeling,

simulation, and model-based drug development. CPT Pharma-

cometrics Syst Pharmacol. 2012;1:e6.

38. US Food and Drug Administration. Center for Drug Evaluation

and Research: pharmacology review: Prezista. Maryland: Silver

spring; 2009.

39. Yau WP, Mitchell AA, Lin KJ, Werler MM, Hernandez-Diaz S.

Use of decongestants during pregnancy and the risk of birth

defects. Am J Epidemiol. 2013;178(2):198–208.

40. Staud F, Cerveny L, Ceckova M. Pharmacotherapy in pregnancy;

effect of ABC and SLC transporters on drug transport across the

placenta and fetal drug exposure. J Drug Target.

2012;20(9):736–63.

41. Joshi AA, Vaidya SS, St-Pierre MV, Mikheev AM, Desino KE,

Nyandege AN, et al. Placental ABC transporters: biological

impact and pharmaceutical significance. Pharm Res.

2016;33(12):2847–78.

42. Schroder H, Leichtweiss HP, Rachor D. Passive exchange and the

distribution of flows in the isolated human placenta. Contrib

Gynecol Obstet. 1985;13:106–13.

43. McCormack SA, Best BM. Protecting the fetus against HIV

infection: a systematic review of placental transfer of antiretro-

virals. Clin Pharmacokinet. 2014;53(11):989–1004.

44. Smith GC, Cameron AD. Estimating human fetal blood volume

on the basis of gestational age and fetal abdominal circumference.

BJOG. 2002;109(6):721–2.

45. Syme MR, Paxton JW, Keelan JA. Drug transfer and metabolism

by the human placenta. Clin Pharmacokinet. 2004;43(8):

487–514.

716 S. Schalkwijk et al.